Obesity, renal hyperfiltration and glomerular filtration rate decline in the general population

(1)

Faculty of Health Sciences Institute of Clinical Medicine

Obesity, renal hyperfiltration and glomerular filtration rate decline in the general population

Results from the Renal Iohexol Clearance Survey

—

Vidar Tor Nyborg Stefansson

A dissertation for the degree of Philosophiae Doctor – April 2019

(2)

List of Tables

Table 1. BMI, waist circumference and waist-hip ratio categories for European, African and Middle Eastern populations according to World Health Organisation and International

Diabetes Federation criteria³. ... 4 Table 2. Criteria for the metabolic syndrome: Three out of 5 criteria must be fulfilled for diagnosis. ... 6

List of Figures

Figure 1. Classification (staging) of CKD by estimated GFR and ACR, 2012 KDIGO

guidelines, kdigo.org ... 12 Figure 2. Flowchart presenting the RENIS study population selection. ... 22

(6)

Acknowledgements

This thesis is based on data from the RENIS study, which was conducted in two rounds at the Clinical Research Unit at the University Hospital of North Norway, and funded by the North Norway Health Authority and a grant from Boehringer-Ingelheim. I received a PhD stipend from UiT The Arctic University of Norway. Without these institutions and their staff, and the people of Tromsø who participated in the RENIS study, this thesis would not have been possible.

First and foremost, I would like to thank my main supervisor Bjørn Odvar Eriksen. You have guided and supported me throughout the long journey that ended with this thesis, and it would not have been possible without your help. Your friendly, optimistic and patient demeanour remained impressively intact despite repeated delays and missed deadlines on my part. I always felt welcome in your office whenever I came knocking, pre-arranged or unannounced.

You are incredibly knowledgeable and a great teacher, and your honest feedback and constructive criticisms invariably raised the quality of all my work.

I would also like to thank my co-supervisors Toralf Melsom and Trond Geir Jenssen for your valuable contributions to the thesis and its constituent papers, and a special thanks to Toralf for including me as the second author of the albuminuria manuscript (paper 3 of this thesis). I would also like to thank Jørgen Schei and Marit Dahl Solbu for your contributions to my manuscripts, and the other friendly people in the Metabolic and Renal Research group.

Finally, I would like to thank my family for supporting and encouraging me through the ups and downs of the last four years. Marie, you are my best friend and the love of my life. Your support is always deeply appreciated, and I would not be where I am today without you.

(7)

List of presented papers

This thesis is based on the following papers:

1. Stefansson VTN, Schei J, Jenssen TG, Melsom T, Eriksen BO: Central obesity associates with renal hyperfiltration in the non-diabetic general population: a cross- sectional study. BMC Nephrology 2016 Nov 10;17(1):172.

2. Stefansson VTN, Schei J, Solbu MD, Jenssen TG, Melsom T, Eriksen BO: Metabolic syndrome but not obesity measures are risk factors for accelerated age-related

glomerular filtration rate decline in the general population. Kidney International 2018 May;93(5):1183-1190.

3. Melsom T, Stefansson VTN, Schei J, Solbu MD, Jenssen TG, Wilsgaard T, Eriksen BO: Association of Increasing GFR with Change in Albuminuria in the General Population. Clinical Journal of American Society of Nephrology 2016 Dec 7;11(12):2186-2194.

(8)

List of abbreviations

ACR Albumin-creatinine ratio BMI Body mass index

BSA Body surface area CKD Chronic kidney disease

CKD-EPI Chronic Kidney Disease Epidemiology Collaboration ΔACR Change in albumin-creatinine ratio

ΔGFR Change in glomerular filtration rate eGFR Estimated glomerular filtration rate ESRD End-stage renal disease

GFR Glomerular filtration rate

KDIGO Kidney Disease: Improving Global Outcomes MetS Metabolic syndrome

mGFR Measured glomerular filtration rate RENIS The Renal Iohexol Clearance Survey

RENIS-FU The Renal Iohexol Clearance Survey Follow-Up RENIS-T6 The Renal Iohexol Clearance Survey in Tromsø 6

RR Relative risk

WC Waist circumference WHO World Health Organization WHR Waist-hip ratio

(9)

Summary

Obesity is a well-known risk factor for several severe diseases, including diabetes and cardiovascular disease. The metabolic syndrome is a concept related to obesity which includes additional risk factors for disease: increased waist circumference, high blood

pressure, elevated fasting glucose, elevated triglycerides and lowered high-density lipoprotein cholesterol levels.

Both obesity and the metabolic syndrome are known risk factors for chronic kidney disease and end-stage renal disease, but their effect on kidney function before reaching those disease states is less clear. The results from previous studies on these subjects are divergent and inconclusive.

The concept of hyperfiltration, a state of elevated GFR (glomerular filtration rate, a measure of kidney function), may contribute to the inconsistency of research results on the subject.

Hyperfiltration is present in diabetes, obesity and hypertension, and is a state of distress which may cause kidney damage in the long term. In the short and medium term, however, it may present as higher or increasing GFR. It may also cause albuminuria, which is an early marker of endothelial damage.

In this thesis, the association between obesity, the metabolic syndrome, changes in GFR and hyperfiltration were explored in the population-based Renal Iohexol Clearance Survey. GFR was measured with an accurate method (iohexol clearance) in 1627 persons in 2007-09 and repeated in 1324 of the same persons in 2013-15. The relationship between changes in GFR and changes in albuminuria was also explored, to further explore the concept of

hyperfiltration as an increase in GFR over time.

We found that obesity was associated with hyperfiltration, but not with accelerated GFR decline. Increased albuminuria was associated with increased GFR. The metabolic syndrome was associated with accelerated GFR decline. The results point to hyperfiltration as an

(10)

important factor in the relationship between obesity and GFR, and that hyperfiltration is associated with albuminuria.

(11)

1 Background

1.1 Obesity

1.1.1 Prevalence

Obesity is a growing problem globally. In large population surveys, body mass index (BMI, defined as body weight in kilos divided by height in metres squared) is the most commonly used measure to define obesity. The World Health Organisation (WHO) criteria classify a person with a BMI ≥25 kg/m² as overweight, while a person with a BMI ≥30 kg/m²is considered obese. In Norway, the estimated prevalence of overweight and obesity increased from 34.0% and 6.7%, respectively, in 1975 to 58.9% and 23.2%, respectively, in 2014¹. In 2016, the Center for Disease Control National Center for Health Statistics estimated the prevalence of obesity in the United States to be 39.8%². In the same year, WHO estimated that more than 1.9 billion people worldwide were overweight, of whom more than 650 million people were obese³.

1.1.2 Obesity as a risk factor

The 2016 Global Burden of Disease Study ranked a high BMI as the second greatest risk factor for global disability-adjusted life years lost in women, and the sixth greatest in men⁴. An estimated 4 million deaths were attributable to high BMI in 2015 globally, and an

estimated 120 million disability-adjusted life years lost⁵. The increased mortality in obesity is largely due to the increased risk of cardiovascular disease and diabetes, but obesity also increases the risk of several other diseases and conditions, including certain cancers, sleep apnoea, infertility and venous thromboembolism^5-7. Compared to persons with a BMI of 20- 25 kg/m², overweight persons have a more than 1.5 times higher prevalence of cardiovascular

(12)

disease and diabetes, while those with a BMI ≥30 kg/m²have a more than 3.5 times higher prevalence of diabetes and hypertension⁸.

1.1.3 Obesity measurements

Obesity is most commonly measured using BMI. Its origin is with the Belgian scientist Adolphe Quetelet, who first used the formula in 1832⁹, but its modern name came from Keys et al. in 1972¹⁰. BMI is calculated from weight divided by height squared, so by definition it does not account for body shape or composition. However, it has become the leading method for the measurement of obesity in large populations due to its simplicity and almost universal availability.

Several simple body measurement techniques other than the BMI have been proposed and used in population studies. Of these, waist circumference (WC) and its ratio to hip

circumference, the waist-to-hip ratio (WHR), are the most commonly used. They capture different aspects of obesity than BMI in that they reflect the placement and distribution of mass in the body rather than the body weight itself. Several studies have pointed to these variables as better predictors of cardiovascular and diabetes risk than BMI^11-14. However, large meta-analyses have not found clinically significant differences between WC, WHR and BMI in predicting these diseases^{15, 16}. WC and WHR are still commonly used in population studies as a supplement to BMI.

BMI, WC and WHR cannot be used to measure the actual amount of fat tissue in the body.

Accurate measurements of fat mass include bioelectrical impedance, dual X-ray

absorptiometry, computed tomography and magnetic resonance imaging. These methods (with the exception of bioelectrical impedance) allow for distinction between visceral (abdominal) fat, subcutaneous fat, and other contributors to body mass such as muscle and bone. Visceral fat, but not subcutaneous fat, has been associated with an increased risk of

(13)

myocardial infarction¹⁷, type 2 diabetes¹⁸, incident chronic kidney disease¹⁹ and metabolic syndrome²⁰, independent of BMI.

1.1.4 Categorisation

Categorising continuous variables may be useful for the purposes of diagnosis, clinical decision making, public policy and communication to the public. WHO has standardised the categorisation of obesity measurements, as presented in Table 1³.

The categories of BMI are fairly well rooted in mortality risk. A large meta-analysis found the lowest mortality in the BMI range of 22.5–25.0 kg/m², with the excess mortality risk

increasing rapidly above a BMI of 30 kg/m^{2 21}. It should be noted that while the BMI

categories are often used universally, several studies suggest a different categorisation should be used for Asian population groups due to the higher risk of diabetes at a lower BMI than in European or African populations²².

The categories of WC are based on a British cohort in a 1995 study by Lean et al.; the cut-off values were chosen based on their ability to identify participants with a high BMI and/or high WHR with high sensitivity and specificity²³. The lower threshold identified subjects with a BMI ≥25 kg/m²and the higher threshold identified subjects with a BMI ≥30 kg/m². The same WC cut-off points were used as one of the five criteria of the metabolic syndrome, which will be covered in more detail in the next chapter of this thesis. The origin of the WHR categories appears to be a 1999 WHO consultation on diabetes, in which the WHR cut-off points were suggested as a criterion for the metabolic syndrome²⁴. The authors offered no source or explanation for this choice of cut-off points, and neither the WC nor the WHR cut-off points appear to be rooted in epidemiological studies of mortality or morbidity.

(14)

Table 1. BMI, waist circumference and waist-hip ratio categories for European, African and Middle Eastern populations according to World Health Organisation and International Diabetes Federation criteria³.

World Health Organisation category Measurement range

Body mass index

Underweight <18.5 kg/m²

Normal 18.5–24.9 kg/m²

Overweight 25.0–29.9 kg/m²

Obesity ≥30.0 kg/m²

Waist circumference

Normal

<80 cm (female)

<94 cm (male)

Increased risk of metabolic complications

≥80 cm (female)

≥94 cm (male)

Severely increased risk of metabolic complications

≥88 cm (female)

≥102 cm (male)

Waist-hip ratio

Normal <0.85 (female)

<0.90 (male)

Severely increased risk of metabolic complications

≥0.85 (female)

≥0.90 (male)

(15)

1.2 Metabolic syndrome

1.2.1 Definition

The concept of metabolic syndrome (MetS) stems from the long-known observation that obesity and a cluster of interrelated risk factors increase the risk for diseases such as diabetes and cardiovascular disease²⁵. Many scientists have contributed to our understanding of the relationships between the various risk factors and diseases, but Reaven is often credited for the modern understanding of the syndrome, with insulin resistance as a core concept²⁶.

The currently used criteria for the syndrome were harmonized in 2009 from different definitions stemming from the WHO and the 2001 National Cholesterol Education Program Adult Treatment Panel III, respectively²⁷. The thresholds for criteria are based on the

diagnostic criteria for hypertriglyceridaemia (high triglycerides), hypoalphalipoproteinaemia (low high-density lipoprotein cholesterol), and pre-diabetes (high glucose), as well as the blood pressure treatment thresholds in diabetes, and the previously mentioned waist circumference thresholds. There is not yet a consensus on which waist circumference threshold should be used for MetS, and both are often presented in studies of MetS. The criteria are listed in Table 2²⁷.

(16)

Table 2. Criteria for the metabolic syndrome: Three out of 5 criteria must be fulfilled for diagnosis.

Category Criteria

Abdominal obesity

Waist circumference: In the United States and Europe, two different thresholds are currently in use by researchers:

≥80 cm (female), ≥94 cm (male) (strict definition)

≥88 cm (female), ≥102 cm (male) (less strict definition)

Other thresholds may apply to different ethnic groups

Elevated blood pressure

Systolic blood pressure ≥130 mm Hg and/or diastolic blood pressure ≥85 mm Hg, or use of antihypertensive medication

Impaired glucose tolerance

Fasting glucose ≥5.6 mmol/L, or use of anti-diabetic medication

High triglycerides levels

Fasting triglycerides ≥1.7 mmol/L, or use of triglycerides- lowering medication

Low high-density lipoprotein cholesterol levels

Fasting high-density lipoprotein cholesterol < 1.29 mmol/L (female), <1.03 mmol/L (male), or the use of cholesterol- altering medication

(17)

1.2.2 Prevalence and relevance

The prevalence of MetS in the United States has increased in tandem with the obesity epidemic, from 22.7% of adults in 1988–94 to 34.2% in 2007-12 in the National Health and Nutrition Examination Surveys, using the less strict definition of MetS^{28, 29}. By the same definition, 25.9% of participants in the North Trøndelag Health Study had MetS in 1995–

97³⁰and 25.5% had MetS in the 2007–8 Tromsø Study³¹. MetS is associated with increased risk of diabetes (relative risk (RR): 3.0)³², cardiovascular disease (RR: 2.4)³³, chronic kidney disease (estimated glomerular filtration rate <60 ml/min/1.73m²; RR: 2.5)³⁴, various cancers³⁵ and all-cause mortality (RR: 1.6)³³.

1.2.3 Utility and controversy of the metabolic syndrome

Critics of the concept argue that these associations do not yield much value because metabolic syndrome is composed of several well-known risk factors and does not necessarily provide additional risk information beyond its constituent components³⁶. Proponents see it as a useful concept to alert and educate patients, healthcare providers and the general public about the high and interrelated risks of insulin resistance, obesity, hypertension and dyslipidaemia which affect a large segment of the population³⁷.

There has also been debate on whether obesity without MetS may constitute a separate, lower-risk “metabolically healthy” form of obesity, in which the risks normally associated with obesity are absent or greatly reduced^{38, 39}. However, obesity appears to increase disease risk significantly even without MetS, although the risk is even higher when MetS is present ^40-

42.

(18)

1.3 Kidney function and albuminuria

1.3.1 Nephron number

The functional unit of the kidney is the nephron. The number of nephrons in humans is set around birth and does not increase afterwards. There is a large variation between individuals in the number of nephrons present in a kidney; estimates vary between 200,000 to more than 2,000,000 per kidney⁴³. Low birth weight and family history of end-stage renal disease are risk factors associated with a lower nephron number^43-46. Adult height and sex are also associated with nephron number and are more easily available to researchers and clinicians than birth weight: tall persons and males generally have higher nephron numbers, though Denic et al. found no sex difference in a multivariable adjusted regression analysis of kidney donors⁴⁷. Height and sex are also associated with birth weight⁴⁸, and these three variables may be seen surrogates for the complex interplay between the genetic, nutritional and intrauterine conditions which determine nephron number43-45, 49, 50. Low nephron numbers have been associated with hypertension and chronic kidney disease^{51, 52}. Nephron numbers decrease gradually with age as nephrons develop sclerosis and cease functioning; donors aged 70-75 years old had 48% fewer nephrons than those aged 18-29 in the study by Denic et al.⁴⁷.

1.3.2 The glomerular filtration rate

Kidney function is usually assessed as the glomerular filtration rate (GFR), which is the total volume of blood filtered through all the glomeruli in the nephrons in both kidneys per minute, expressed as ml/min. By tradition, this whole-kidney GFR is adjusted for 1.73 m² of body surface area (BSA) to reduce the spread in GFR seen in people of different sizes, although this is not without controversy, as we will see in chapter 1.3.3.

In everyday clinical practice and in most population studies, GFR is estimated using the serum concentration of creatinine or cystatin C and an estimation equation. There are several

(19)

equations for estimated GFR (eGFR), but the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation is currently the most commonly used equation for adults⁵³, and the Schwartz equation is used for children⁵⁴. The Berlin Initiative Study 1 equation has been developed for the elderly,⁵⁵ but there is no general agreement about which equation is best for this age group. A unified equation for all ages has been proposed, but has not yet been widely adopted⁵⁶.

1.3.3 Critiques of eGFR and body surface area standardisation

All eGFR equations are hampered by the fact that they are estimates based on serum levels of creatinine and/or cystatin C, rather than actual measurements of GFR. Around 10–20% of CKD-EPI eGFR values differ by more than 30% from the measured GFR (mGFR) value, and the absolute differences (in ml/min/1.73m²) between eGFR and mGFR are larger in the higher ranges of GFR^{53, 57}. Additionally, the serum concentrations of both creatinine and cystatin C are known to be influenced by non-GFR-related factors such as muscle mass, cardiovascular risk factors and inflammation, which lead to biased eGFR estimates and contribute to the imprecision of eGFR^58-62.

Another critique points to the traditional standardisation of GFR to a body surface area (BSA) of 1.73m². The equation for estimating BSA was created by the Du Bois brothers in 1916⁶³. The choice of 1.73 m² is based on the average BSA measured in volunteers in 1925⁶⁴; the average BSA today is significantly higher. One strand of the critique rejects the basis for any standardisation at all, because it leads to an underestimation of GFR in obese persons in particular^65-67. A person who gains weight will have higher GFR due to increased metabolic needs of the heavier body, but does not grow any new nephrons to handle this task, so GFR per nephron (single-nephron GFR, see chapter 1.5) will increase. The corresponding increase in BSA, however, apparently mitigates some of the GFR increase if one adheres to the

(20)

tradition of BSA standardisation. Even if the premise of standardisation of GFR is accepted, BSA is a poor choice to improve comparability of GFR across body sizes. Other variables such as extracellular fluid or total body water have been proposed as better alternatives, but have not been widely used in research or clinical practice^{68, 69}.

1.3.4 Measuring GFR

One way to allay the problems associated with eGFR is measuring GFR precisely with an exogenous marker. The gold standard is continuous infusion of the inert marker inulin, but it is cumbersome and expensive to use in practice. The contrast substrates iohexol and

iothalamate are easier to use, and correlate very well with inulin clearance^70-72.

All GFR measurement options require the injection or infusion of the marker followed by one or multiple measurements of the marker concentration in blood or urine. Unfortunately, this causes the methods to be costly and more time-consuming than the serum-based estimates and are thus regarded by many as unfeasible for everyday clinical practice.

mGFR is mostly used in settings where the precise GFR of a patient is central to the decision to treat or not to treat, or when evaluating whether potential kidney donors are eligible to donate a kidney. The Renal Iohexol Clearance Survey (RENIS) study, presented in detail in chapter 3.1, is the only large general population study using repeated GFR measurements over time.

1.3.5 Albuminuria

Albumin is a protein present in the blood, and is normally excreted in very low quantities in the urine in healthy individuals. The presence of elevated albumin levels in urine is termed albuminuria, and is interpreted as a marker of kidney disease and endothelial dysfunction. It is often measured using immunoturbidometric assays, but high albumin levels can also be

(21)

detected as proteinuria with a standard dipstick test. Both albumin and creatinine

concentrations are measured (in mg and mmol, respectively), and albumin is standardised to creatinine as the albumin/creatinine ratio (ACR), which has a high correlation with the albumin excretion rate⁷³. Persistent albuminuria (>3 months of ACR >3 mg/mmol) is sufficient for a chronic kidney disease (CKD) diagnosis independent of GFR⁷⁴, and is a marker of increased risk at any CKD stage (see chapter 1.4.1).

Albuminuria is associated with increased risk of severe CKD, cardiovascular disease and mortality, even at excretions lower than the currently used Kidney Disease: Improving Kidney Outcomes (KDIGO) standard of 3 mg/mmol^{75, 76}. Both obesity and MetS are

associated with increased albuminuria^77-80. A recent report from the RENIS cohort found that even trace amounts of albuminuria were associated with more rapid subsequent mGFR decline⁸¹.

1.4 Chronic kidney disease

1.4.1 Definitions

CKD is a prolonged state of reduced kidney function or kidney damage caused by a variety of diseases and risk factors. It was defined by the Kidney Disease Outcomes Quality Initiative study group in 2002 and further refined in 2012 by the KDIGO study group^{74, 82}. According to the criteria, CKD is defined by a reduced eGFR and/or increased ACR for more than 3

months. The cause of kidney disease is also formally a part of the definition, but does not seem to play an important role in the practical staging of CKD. The KDIGO classification of CKD is tabulated in Figure 1⁷⁴.

(22)

Figure 1. Classification (staging) of CKD by estimated GFR and ACR, 2012 KDIGO guidelines, kdigo.org

The KDIGO CKD criteria have been criticised for not taking age into account. GFR declines slowly when people get older, and the prevalence of CKD thus increases rapidly with age, approaching 50% among persons aged over 70⁸³. Most elderly persons diagnosed with CKD have an eGFR between 45 and 59 without albuminuria, and their prognosis is good⁸⁴. Others argue that introducing age into the classification would confuse patients and healthcare workers, and that GFR-related drug dosage restrictions should be based on GFR regardless of age⁸⁵.

(23)

1.4.2 Incidence and prevalence of CKD

End-stage renal disease (ESRD) is the most severe end stage (stage G5) of CKD. The

incidence of ESRD in the US rose from a standardised incidence rate (standardised to the age, sex and race distribution of the US in 2011) of 87 per million in 1980 to 357 per million in 2015, however the standardised incidence rate seems to have plateaued in the last 5-10 years⁸⁶.

The prevalence of all stages of CKD increased significantly in the United States in the 1980s and 90s, including both the moderate, usually asymptomatic stages (stages G1-3, Figure 1) and ESRD, but has remained fairly stable in the last two decades⁸⁶. ESRD is a serious condition, which requires intrusive and costly renal replacement therapy (dialysis or

transplantation). Dialysis has a very high mortality rate of 164 deaths per 1000 patient-years (age-sex-race-standardised), while transplant recipients have a much lower rate (29 per 1000 patient-years). However, standardised death rates for both dialysis and transplant recipients declined by 29% and 40%, respectively, between 2001–16. In 2011–14, an estimated 14.8%

of the adult United States population had CKD, including 0.2% with ESRD⁸⁶.

Globally, estimated deaths attributable to CKD rose from 937,700 in 2005 to 1,234,900 in 2015 according to the Global Burden of Disease survey⁸⁷. CKD attributable to hypertension was the largest driver of deaths, followed by diabetes.

In Norway, the incidence of ESRD has stabilised, but the prevalence still increases because of better survival among those with ESRD⁸⁸. In 2016, 554 persons (105.8 per million) began renal replacement therapy. The three main causes were vascular/hypertensive (34%), glomerulonephritis (17%) and diabetic nephropathy (16%). This suggests that a significant proportion of ESRD may be preventable because it is rooted in diseases that can be

effectively treated or prevented by modifying lifestyle-associated risk factors. The prevalence

(24)

of persons in renal replacement therapy in Norway was 4969, or 948.9 per million inhabitants⁸⁸.

1.5 Kidney physiology in hyperfiltration and ageing

1.5.1 Hyperfiltration

1.5.1.1 Background

While CKD is defined by a low GFR or kidney damage, a high GFR is not necessarily a sign of health. The theory of hyperfiltration, proposed by Brenner et al. in 1996 based on research from the 1980s and 90s⁸⁹, considers GFR at the single-nephron level (GFR per nephron, or whole-kidney GFR divided by nephron number). Brenner et.al demonstrated that rats with a reduced nephron number had high single-nephron GFR (hyperfiltration) and/or higher intraglomerular pressure. This state of hyperfiltration was in turn associated with podocyte damage, mesangial expansion, albuminuria and finally glomerulosclerosis and GFR decline⁹⁰. Nephron loss may further increase the stress in remaining nephrons, causing a vicious cycle.

Hyperfiltration has been shown to occur in diabetes⁹¹, but Brenner et al. proposed it as a general mechanism behind many diseases or conditions with nephron loss, including hypertension and obesity⁹². In a recent study, Melsom et al. found that high mGFR adjusted height, sex and age predicted faster subsequent mGFR decline in two cohorts of different ethnic origin: Pima Indians with diabetes and the Norwegian RENIS cohort¹¹².

1.5.1.2 Mechanisms

The mechanisms of hyperfiltration are not fully understood, but may include several

interacting factors. Vascular resistance in the afferent and efferent arterioles regulates single- nephron GFR. The antihypertensive drugs angiotensin converting enzyme inhibitors and angiotensin II receptor blockers both target angiotensin II and cause efferent arteriole vasodilation⁹³. They have been shown to reduce GFR in the short term, but to slow GFR

(25)

decline in the longer term, possibly because of a reduction of hyperfiltration⁹⁴. A similar pattern of reduced hyperfiltration is seen with the use of sodium-glucose cotransporter 2 inhibitors⁹⁵. These drugs block glucose reabsorption in the proximal tubule, causing the glucose to be excreted in the urine rather than reabsorbed. Because the blocked cotransporter also transports natrium (sodium), natrium reabsorption in the proximal tubule is reduced as well, increasing natrium concentrations in the macula densa. This causes afferent arteriole vasoconstriction by tubuloglomerular feedback, which is mediated by an increase in

adenosine. Interestingly, a recent study found increased adenosine concentrations in the urine of patients with type 1 diabetes who were treated with the drug empagliflozin⁹⁶. Other

potential mechanisms involved in hyperfiltration include intrarenal nitric oxide signalling and mechanical stress from glomerular hypertension^97-100.

1.5.1.3 Epidemiology

While creatinine-based eGFR is a poor method for hyperfiltration research due to its inaccuracy, large longitudinal population studies using the method have shown increased mortality and morbidity in persons with high creatinine-based eGFR⁵³. Traditionally, this has been explained as a falsely elevated eGFR from low serum creatinine being the result of low muscle mass due to wasting from chronic disease, such as cancer or severe emphysema.

However, studies that have measured muscle mass and accounted for concurrent disease still found higher mortality in hyperfiltration¹⁰¹.

Several studies suggest associations between hyperfiltration and many well-known ESRD risk factors, including pre-diabetes and diabetes, hypertension, obesity, albuminuria and

smoking92, 101-108. Some interventions which reverse these factors, such as treatment of hypertension with losartan, treatment of diabetes with sodium glucose co-transporter 2-

(26)

inhibitors and significant weight loss after gastric bypass surgery, result in a GFR decrease which may represent the normalisation of previous hyperfiltration95, 109-111

Obesity, MetS and hyperfiltration have only been studied to a limited extent, with divergent results. These conditions are by their nature particularly affected by the imprecision of eGFR because of altered muscle mass in obesity, and low-grade systemic inflammation in obesity and MetS. The customary adjustment for 1.73 m² of BSA also distorts the estimation of hyperfiltration in the obese. These factors may explain the divergent results among different study populations69, 103, 113-121, and are discussed in detail in papers 1 and 2 in this thesis.

1.5.1.4 Definition

There is no consensus on a common definition of hyperfiltration based on whole-kidney GFR.

Some have used an arbitrary, round GFR cut-off value¹⁰³, while others have used percentiles in their study population57, 92, 104, 114. However, a simple absolute or BSA-adjusted whole- kidney GFR cut-off does not adequately consider the varied nephron endowment of

individuals, and is less likely to reflect single-nephron GFR. A meta-analysis by Chagnac et al. suggested that a common hyperfiltration definition should at the very least involve an adjustment of GFR for age and gender to at least partially account for nephron numbers¹²². However, the authors did not suggest a definition themselves. A recent study by Chakkera et al. compared different methods to establish hyperfiltration definitions more consistent with single-nephron GFR in kidney donors. The study is discussed in detail in chapter 5.2.4 of this thesis.

1.5.2 The ageing kidney and GFR decline

As mentioned in chapter 1.3.1, the number of nephrons declines gradually with age, as they cease to function and the glomeruli become sclerotic47, 123-125. Other changes to the kidneys in old age include increased atherosclerosis in renal vasculature, interstitial fibrosis, tubular

(27)

necrosis, an increased number of renal cysts, and a reduction in cortical volume¹²⁴. The nephron loss causes whole-kidney GFR to decline slowly with age, while single-nephron GFR remains fairly constant with age in healthy kidney donors, except in the oldest age group (>75 years old)¹²⁶.

In longitudinal studies of whole-kidney GFR, the rate of decline varies greatly. Some risk factors for CKD and ESRD may attenuate GFR decline, or even increase GFR in the short- term, but still increase the risk of ESRD in the long-term81, 127-130. This apparent paradox may in part be due to the negative effects of hyperfiltration.

Because no new nephrons are created after birth, an increase in GFR must represent an increase in single-nephron GFR, while a decrease in GFR may be due to a lower single- nephron GFR, a loss of nephrons, or both. This suggests a possible alternative definition of hyperfiltration: a significant increase in whole-kidney GFR over time.

1.6 Obesity as a risk factor for CKD

Obesity is known to increase risk of diabetes, hypertension and cardiovascular disease, and also CKD and ESRD^131-135. While a large part of the association with CKD is due to the first three factors, obesity may also be associated with CKD and ESRD independent of these intermediaries³⁹. In a large meta-analysis, Hsu et al. found a relative risk for ESRD ranging from 1.9 for those with a BMI from 25–29.9 kg/m² to 7.1 in subjects with a BMI ≥40 kg/m²

134.

However, the relationship between obesity, hyperfiltration and GFR decline in the general population is less clear, with conflicting study results19, 69, 103, 136, 137. The inconsistency of the results may be due to a combination of the inherent inaccuracies of eGFR, the misleading BSA correction of GFR, and the nature of hyperfiltration. As we explored in the previous

(28)

chapter, hyperfiltration may cause GFR to stabilize or increase in the short term, concealing detrimental effects on the kidney because eGFR appears to be normal. The damage to the kidneys may be reflected in lower eGFR at a much later stage, when preventative efforts may be less effective. A distinct form of kidney damage from severe obesity is obesity-related glomerulopathy, likely related to obesity-related hyperfiltration¹³⁸.

In summary, obesity, MetS and CKD are widespread globally, and are major causes of shortened lifespans and decreased quality of life. While obesity and the metabolic syndrome have consistently been shown to increase the risk of ESRD in epidemiological studies with long follow-up periods, their relationship with hyperfiltration and the age-related decline in GFR are not very well understood. This is in large part because of methodological problems caused by the use of eGFR instead of actual GFR measurements.

2 Aims

The primary aim of this thesis was to explore the relationship between obesity, metabolic syndrome, hyperfiltration, and the subsequent GFR decline rate. Since albuminuria is an important early sign of kidney dysfunction, we also examined the association of

hyperfiltration (defined as an increase in GFR) with an increase in albuminuria.

(29)

3 Methods

3.1 Participants

3.1.1 RENIS-T6

RENIS began in 2007 as a sub-study of the 6^th Tromsø study¹³⁹. The purpose of the study was to measure GFR with an accurate method in a large population of fairly healthy participants, representative of the general population. The chosen age group, 50–62 years old, was chosen because many people of that age have risk factors for lifestyle-associated diseases such as diabetes, chronic kidney disease and cardiovascular disease, but are still fairly healthy and have yet to develop those diseases. By studying them during ageing, it is possible to see which factors influence the course of GFR over time.

The 6^th Tromsø study invited all citizens of Tromsø 60–62 years of age, and a random sample of 40% of those aged 50-59 years old¹⁴⁰. This amounted to 5464 people, of whom 3564 (65%) completed both rounds of the main part of the study.

The exclusion criteria for the first round of the RENIS study, named RENIS-T6, were diabetes, any renal disease except urinary tract infections, angina pectoris, myocardial infarction or stroke. Overall, 739 of those who completed the Tromsø study were excluded, leaving 2825 eligible people who were invited to RENIS. Of these, 2107 responded

positively, but a further 125 were ultimately excluded because they reconsidered and withdrew, reported allergic reactions to iodine or latex or for other practical reasons. The selection process for the study population of RENIS is also shown in Figure 2.

The predetermined target study population for RENIS-T6, based on power calculations, was 1600. Participants had their appointments scheduled in a random order. When the number of investigations had reached 1632, the study was stopped, leaving the remaining 350 eligible

(30)

potential participants uninvited. Five investigations were technical failures, leaving 1627 as the final study population of RENIS-T6. The investigations took place at the Clinical

Research Unit at the University Hospital of North Norway between November 2007 and June 2009. All participants provided written informed consent to participate, and the Regional Ethics Committee of Northern Norway approved the study. The study was performed in compliance with the Declaration of Helsinki.

3.1.2 RENIS-FU

The second round of RENIS, RENIS-FU (Follow-Up), invited all participants from RENIS- T6 to repeat the protocol between September 2013 and January 2015, except those who had died (n=23) and 7 individuals who had a possible allergic reaction to iohexol. A total of 1324 participants (83% of those eligible) attended the follow-up. Eighty-eight participants were randomly selected to undergo two GFR measurements with a median (interquartile range) 35 (22–49) days between the measurements, resulting in 3 total measurements for these persons.

The extra measurement allowed for the estimation of an intra-individual variation coefficient.

A flowchart for the study population selection for each paper is presented in Figure 2.

3.1.3 Study population selection

For the first paper of this thesis, the study population included all participants from RENIS- T6, except those who had previously unknown diabetes (fasting plasma glucose ≥7.0 mmol/L and/or haemoglobin A1c ≥6.5%) and those who lacked waist or hip circumference

measurements, leaving 1555 participants. For the second paper, only participants who participated in both RENIS-T6 and RENIS-FU were included (n=1324). The same exclusion criteria as in the first paper were applied, with the additional exclusion of two participants who lacked triglyceride measurements at baseline, leaving 1261 participants as the study population. In the third paper, participants who participated in both RENIS-T6 and RENIS-

(31)

FU were included, except for those with previously unknown diabetes and those who had albuminuria (ACR >30 mg/g) at baseline, leaving 1246 persons in the study population.

(32)

Figure 2. Flowchart presenting the RENIS study population selection.

N=3564Participated in both rounds the 6thTromsø Study N=5464Invited to the 6thTromsø Study, aged 50-62 years old

N=2825Invited to RENIS-T6

N=1627Participated and successfully completed RENIS-T6

N=1597Invited to RENIS-FU

N=1324Participated and successfully completed RENIS-FU N=1900Did not attend, or attended only the 1stround

N=739Reported cardiovascular disease, diabetes or renal disease

N=848Did not respond or withdrew from the study (n=766), technical failure (n=5), or allergy/other (n=77) N=350Did not participate becausepopulation target was reached N=1555Paper 1 study population N=72Diabetes (n=33) or missing waist circumference measurement (n=39) N=30Died (n=23) or possible allergic reaction (n=7)

N=273Did not respond or failed to show up at appointment (n=268), or technical failure (n=5)

N=63Diabetes at baseline (n=25) or missing waist circumference (n=36) or triglycerides measurement (n=4)

N=1261Paper 2 study population N=78Diabetes (n=25) or albuminuria (n=17) at baseline, or diabetes at follow-up (n=36)

N=1246Paper 3 study population

(33)

3.1.4 Participant instructions and body measurements

All subjects showed up at their morning appointment (between 8 and 10 AM) after an overnight fast, having abstained from smoking for 12 hours. They were instructed not to eat unusually large portions of meat or take non-steroidal anti-inflammatory drugs during the last two days before the examination and to drink two glasses of water before arrival. Upon arrival, their height and weight were measured, as were their waist and hip circumferences.

They answered a large questionnaire at home before arrival, which included questions on tobacco and alcohol use, medical history and drug use. A study nurse re-examined the questionnaire with the participants upon arrival, including a thorough review of all current medication usage and medical history to reduce the risk of misclassification

The protocol was the same for both rounds of RENIS, with the exception that in RENIS-T6 the questionnaire, ACR, haemoglobin A1C, waist and hip circumferences and height were measured as part of the 6^th Tromsø study, a median (interquartile range) 5.2 (3.0–6.2) months before RENIS-T6.

Body weight was measured to the nearest 0.1 kg on a digital scale. Height was measured with a wall-mounted measuring tape to the nearest centimetre. Waist circumference was measured horizontally over the umbilicus at the point of expiration. Hip circumference was measured around the greatest protrusion of the buttocks.

3.2 Laboratory measurements

3.2.1 Albuminuria measurements

Participants collected fasting morning samples of urine the last two days before the

examination, and on the morning of the examination. Albumin and creatinine were measured in fresh (unfrozen) specimen, and the median ACR from the measurements for each

(34)

individual was used for the studies. The creatinine concentration was measured using colorimetric methods (Jaffe’s reaction), while albumin was measured using the immunoturbidimetric method. Both were done on an ABX Pentra Micro-albumin CP autoanalyser. Although the limit of detection of the assay is given as 4 mg/L in the

documentation, the PENTRA instrument in practice detects albumin concentrations down to 1 mg/L. This had consequences for paper 3, see chapter 4.3.1 of this thesis.

3.2.2 Single-sample iohexol clearance measurements

A Teflon catheter was inserted into the antecubital vein, and blood samples were drawn for analyses. Five millilitres of iohexol (Omnipaque, 300 mg I/ml) was injected, the syringe was weighed before and after injection, and the catheter was flushed with 30 ml of isotonic saline.

After the iohexol injection, participants were served a light breakfast and were free to walk around or relax at will.

After an individually pre-specified period of time, calculated using the Jacobsson’s method based on eGFR from creatinine¹⁴¹, a new blood sample was taken for iohexol analysis. The exact time from iohexol injection to blood sample extraction was measured using a

stopwatch. High performance liquid chromatography was used to measure the iohexol concentration, as described by Nilsson-Ehle¹⁴². The analytic coefficients of variation were 3.0% in RENIS-T6 and 3.1% in RENIS-FU. The mean coefficient of variation for the intra- individual variation in GFR among the 88 participants who had two GFR measurements in RENIS-FU was 4.2%¹²⁷.

3.2.3 Other measurements

Fasting serum glucose, total cholesterol, LDL cholesterol, HDL cholesterol and triglyceride concentrations were measured on a Modular P800 (Roche Diagnostics). The insulin

concentration was measured with an enzyme-linked immunosorbent assay kit (DRG

(35)

Instruments, Marburg, Germany). The intra- and inter-assay coefficients of variation were 4.7% and 6.3%, respectively. Insulin resistance was expressed by the homeostasis model assessment equation, multiplying fasting glucose (mmol/L) by fasting insulin (mU/L) and dividing the result by 22.5¹⁴³. In RENIS-T6, haemoglobin A1c was measured as part of the 6^th Tromsø study using a liquid chromatographic method.

Serum creatinine was measured using an enzymatic assay standardised to the isotope dilution mass spectrometry method (CREA Plus, Roche Diagnostics). Cystatin C was analysed with a particle enhanced turbidimetric immunoassay with reagents from Gentian (Gentian, Moss, Norway) and a Modular E analyser (Roche Diagnostics). The cystatin C measurements were then recalibrated to the international reference standard using a Cobas 8000 (Roche

Diagnostics). CKD-EPI equations were used to estimate GFR⁵³.

Office blood pressure was measured at the study site after two minutes of rest using an automated device (model UA799; A&D, Tokyo, Japan). Daytime ambulatory blood pressure was measured using weighted daytime (10:00–22:00) averages of blood pressure measured at 20-minute intervals. Further details of the blood pressure measurements in RENIS have been described previously¹⁴⁴.

3.3 Statistical methods

3.3.1 Hyperfiltration definition

In paper 1, hyperfiltration was defined using two different approaches. In both cases multiple linear regression models were used, with the natural logarithm (ln) of unadjusted GFR

(mL/min) as the dependent variable. In one definition, age, sex and height (all associated with nephron number) were added as independent variables, while in the second, the same

variables were added along with body weight.

(36)

A participant was defined as having hyperfiltration if the regression residual was greater than the 90th percentile in the distribution of residuals in the regression analyses for the respective hyperfiltration definition. The definitions were exemplified and explained in more detail in the paper. In the third paper, hyperfiltration was defined as an increase in GFR during the follow-up period.

3.3.2 Metabolic syndrome and obesity categorizations

The metabolic syndrome was defined using the previously mentioned harmonised

WHO/International Diabetes Federation definition (chapter 1.2.1 and Table 2 in this thesis).

The obesity categorisations were also in line with the international standards described in chapter 1.1.4 and Table 1 of this thesis.

3.3.3 Descriptive statistics

The study population characteristics in all papers were presented as the mean (standard deviation) values, median (interquartile range) in cases of skewed data, or numbers

(percentages) where appropriate. Differences in characteristics between categories were tested with paired t tests for mean values, Wilcoxon signed rank tests for median values, and

McNemar tests for paired dichotomous variables, respectively.

3.3.4 Regression analyses

In the first paper, the main results were analysed with logistic regression models with hyperfiltration (two different definitions) as the dependent variable. In the second paper, the main results were analysed with linear regression models, with change in absolute

(unadjusted) mGFR between the RENIS-T6 (baseline) and RENIS-FU (follow-up)

measurements as the dependent variable. In the third paper, the main results were analysed in linear regression models with change in ACR between baseline and follow-up as the

dependent variable. Logistic regression models were also used with incident albuminuria

(37)

(ACR >30 mg/g at follow-up) as a dichotomous dependent variable. The analyses were explained in greater detail in the respective papers. All analyses were performed using STATA MP 14 (www.stata.com).

4 Main results

4.1 Paper 1. Central obesity associates with renal

hyperfiltration in the non-diabetic general population: a cross-sectional study

A total of 1555 participants from RENIS-T6 were examined for associations between BMI, WC or WHR and two different definitions of hyperfiltration. The first hyperfiltration definition adjusted for age, sex and height was associated with all three obesity measures in logistic regression models. The associations remained significant after adjustments for potential confounders including the component risk factors of the metabolic syndrome. This definition is an attempt to adjust for some of the main factors known to influence nephron number. When another hyperfiltration definition was used, which adjusted for weight in addition to age, sex and height, only WHR remained significantly associated with hyperfiltration after controlling for confounders. These results suggest that obesity is

associated with hyperfiltration in the general non-diabetic population, even when controlling for obesity-associated potential confounding factors. Furthermore, elevated WHR is

associated with hyperfiltration even when using a weight-adjusted definition.

(38)

4.2 Paper 2. Metabolic syndrome but not obesity measures are risk factors for accelerated age-related glomerular filtration rate decline in the general population

A total of 1261 persons who participated in both RENIS-T6 and RENIS-FU were examined for associations between obesity, MetS and changes in the decline rate of GFR during the mean 5.6 years between the two rounds of RENIS. Obesity, measured with BMI, WC or WHR, was not associated with a statistically significant change in the rate of GFR decline.

MetS, however, was associated with a -0.30 ml/min/year faster GFR decline in multivariable adjusted linear regression models. The triglyceride criterion of MetS was the main driver of this result.

4.3 Paper 3. Association of increasing GFR with change in albuminuria in the general population

The relationship between the change in GFR and the simultaneous change in ACR during the mean 5.6 years between RENIS-T6 (baseline) and RENIS-FU (follow-up) was explored. The change in GFR was termed ΔGFR (defined as GFR at follow-up minus GFR at baseline). A positive ΔGFR signified an increase in GFR, and a negative ΔGFR signified a decline in GFR. The same principles applied to changes in ACR (ΔACR).

There was a positive association between ΔGFR and ΔACR in multivariable adjusted linear regression analyses: ΔACR was 8.4% higher per standard deviation of ΔGFR. When

participants were split into two groups based on those whose GFR increased during the study period (ΔGFR >0, n=343) and those whose GFR declined (ΔGFR <0, n=903), the group whose GFR increased experienced a 16.3% higher ΔACR (see chapter 4.3.1 for additional analyses of these data). When logistic regression was used to find the odds ratio of incident albuminuria (ACR >30 mg/g at follow-up), those with a higher ΔGFR had an odds ratio of 2.13 for incident albuminuria per standard deviation. The group whose GFR increased during

(39)

the study period (ΔGFR >0) had an odds ratio of 4.98 of incident albuminuria compared to those whose GFR declined (ΔGFR <0).

4.3.1 Additional analyses for Paper 3

As mentioned in chapter 3.2.1, urinary albumin in both the Tromsø 6 and RENIS-FU examinations were analysed with the ABX Pentra Micro-albumin CP (Horiba ABX, Montpelier, France). Although the limit of detection of the assay is given as 4 mg/L in the documentation, the PENTRA instrument in practice reports albumin concentrations as low as 1 mg/L. These results were used in paper 3 and in previous publications from the Tromsø Study. In paper 3, all urinary samples with undetectable albumin concentrations (i.e. <1 mg/L) were assigned an ACR of 0.10 mg/mmol, which corresponds to the lowest ACR observed in samples with detectable albumin concentration. Because this is not formally correct, we have now repeated the analyses of the data in paper 3 using two methods: first, we treated all observations of albumin concentrations lower than 4 mg/L as left-censored and repeated the multiple linear regression of ACR in Table 2 of paper 3 using interval regression.

Second, because there is no similar procedure for a dichotomous dependent variable, we used multiple imputation to impute the ACR for observations with albumin below 4 mg/L. This was done by extending the previously described RENIS multiple imputation model to include interval regression for the missing ACRs¹⁴⁵. The 50 imputed datasets were then used to repeat the analyses in Table 3 of paper 3.

For Table 2, the difference in GFR between baseline and follow-up predicted the absolute difference in ACR in all three models (p<0.05). When the difference between log-transformed ACR was used as the dependent variable, the results were similar to the original results, but borderline statistically significant in the fully adjusted model (p=0.05). However, the

dichotomised variable for GFR increase (ΔGFR>0) was not a statistically significant predictor

(40)

of ACR increase. For the analyses with multiple logistic regression in Table 3, the results were similar to the original results in all models (p<0.05).

5 Discussion

5.1 Methodological considerations

5.1.1 Selection bias

Selection bias occurs when the study population and target population (from which the study subjects are recruited) differ with regard to exposures or outcomes of interest. The RENIS cohort was recruited from participants in the 6^th Tromsø Study. In that study, approximately 65% of those invited in the age group 50–62 years old participated. All Tromsø study

participants in that age group were invited to RENIS, except those who had reported diabetes or a history of myocardial infarction, stroke or renal disease. The response rate was 75% for those invited to RENIS. Even though these percentages are high by international standards, they nevertheless leave room for significant selection bias.

We know that those who chose not to participate in the Tromsø Study were more likely to be male and in the younger age group (50–55 years old)¹⁴⁰. We do not know non-participants’

motivations, and can only speculate whether their choice not to participate makes them more or less likely to be obese, have metabolic syndrome, albuminuria, or differ from participants in other ways. A study of non-responders in the North Trøndelag Health Study found that the reasons for non-participation differed across age groups, with younger participants more likely to report being hindered by time constraints, while older participants were more likely to report poor physical health as a reason to not participate, or stated that they already

Obesity, renal hyperfiltration and glomerular filtration rate decline in the general population